Floating‐Gate Synaptic Transistors for Energy‐Efficient Neuromorphic Computing
Nan Zhang, Yi Wang, Yujie Yan, Shujin Chen, Yu Zhang, Changsong Gao, Lingjie Sun, An Xie, Fangxu Yang, Wenping Hu
Abstract
By integrating nonvolatile memory and processing, floating-gate synaptic transistors (FGSTs) have emerged as a pivotal platform for energy-efficient neuromorphic computing, overcoming limitations inherent in conventional Von Neumann architectures. These devices utilize a unique floating-gate layer to facilitate charge storage and manipulation. This review presents a comprehensive overview of recent advancements in FGST device design, focusing on innovative floating-gate structures, diverse floating-gate material systems, and advanced tunneling dielectric layers. These innovations have significantly enhanced synaptic performance, including near-linear conductance modulation, ultralow energy consumption, multilevel storage, extended retention times, and robust endurance characteristics. Consequently, FGSTs achieve remarkable pattern-recognition accuracy and effectively mimic complex biological plasticity rules. Moreover, their integration into neuromorphic sensory systems for vision, audition, touch, and neuronal behavior enables these devices to conduct high-fidelity real-time multimodal and reconfigurable processing. Despite these advancements, challenges persist in scaling synaptic energy to femtojoule levels, enhancing the mechanical flexibility of wearable electronics, improving operational stability, and developing large-scale synaptic devices array. This paper outlines strategic pathways in materials and architecture to steer the development of FGSTs toward highly efficient, brain-inspired neuromorphic hardware.